GPT 5.6 Luna is the smallest tier of OpenAI's GPT-5.6 family, released to general availability on July 9, 2026 alongside gpt-5.6-sol and gpt-5.6-terra. OpenAI designed it for cost-sensitive, high-volume workloads — the tier that roughly replaces the nano models of earlier GPT-5 generations. If you're asking what is gpt 5.6 luna api access good for, the short answer: any job where you process a lot of tokens and the task is easy to verify — classification, tagging, extraction, summarization, and first-line chat.
The openai gpt 5.6 luna api is available on GPTProto with the same model string and OpenAI-compatible schema, billed from one balance that also covers Sol, Terra, and 200+ other models.
| Spec | gpt-5.6-luna |
|---|---|
| Provider | OpenAI |
| Modality | Text & Image in / text out |
| Context window | 1,050,000 tokens |
| Max output | 128,000 tokens |
| Knowledge cutoff | February 16, 2026 |
| Reasoning effort | none / low / medium / high / xhigh / max |
| Structured outputs | Strict JSON schema mode |
| GPTProto price | $0.80 input / $4.80 output per 1M tokens |
| Official price | $1 input / $6 output per 1M tokens |
| Model string | gpt-5.6-luna |
What high-volume work costs on GPTProto ($0.80/$4.80 per 1M):
| Workload | Input | Output | GPTProto cost | Official cost |
|---|---|---|---|---|
| 1,000 document summaries | 2M | 300K | $3.04 | $3.80 |
| Classification batch | 10M | 500K | $10.40 | $13.00 |
| Chatbot, monthly | 50M | 10M | $88.00 | $110.00 |
Difference Between GPT 5.6 Luna and Sol
Luna and Sol are the two ends of the gpt 5.6 luna model family's cost-capability curve. The published numbers:
| gpt-5.6-luna | gpt-5.6-sol | |
|---|---|---|
| Tier | Fastest, cheapest (nano-equivalent) | Flagship |
| Terminal-Bench 2.1 | 84.7% | 88.8% (91.9% with ultra) |
| SWE-Bench Pro | 62.7% | 64.6% |
| Artificial Analysis Intelligence Index | 51 | 59 |
| Artificial Analysis Coding Agent Index | 74.6 | 80 |
| Long-context behavior | Degrades more sharply at depth (OpenAI testing) | Strongest retrieval of the family |
| ultra multi-agent mode | No | Yes |
| Official price per 1M | $1 / $6 | $5 / $30 |
| GPTProto price per 1M | $0.80 / $4.80 | $4 / $24 |
Honest read: Sol is roughly 4–8 points better across benchmarks at five times the price. That trade is worth it for long agent runs and tasks that are expensive to redo; it is rarely worth it for bulk classification or summarization. A common production pattern is routing: default to cheaper gpt 5.6 luna api calls, escalate to Sol when a request fails validation or exceeds a complexity threshold. For the middle option, see gpt-5.6-terra; the full three-tier table lives on the Sol page.
